Koru Voice
Voice typing that never leaves your computer
Koru Voice turns speech into text on your own computer. Hold Ctrl+Shift+D (Cmd+Shift+D on a Mac) and talk, and your words appear where you were typing, in Koru or, on Windows, in any other app. Speech recognition runs locally, your audio never goes to a server, and it is free.
How does voice typing work in Koru?
Hold Ctrl+Shift+D, or Cmd+Shift+D on a Mac, and talk. What you say appears where your cursor is, underlined and dimmed while Koru is still hearing it, and becomes ordinary text when you let go. It works anywhere you can type in Koru: a document, search, a rename box or the find bar.
- One dictation is one undo. Nothing is written while you are still speaking, so a single undo takes back the whole sentence.
- A small chip shows what is happening: Starting while the model loads, Listening once the microphone is open, Finishing after you let go, with a meter that moves with your voice.
- Click the chip to finish and keep the words, instead of letting go of the keys.
- Prefer not to hold a key? Settings can make it a toggle: press once to start, again to stop. Or type
/dictatein a document.
Is my voice sent to the cloud?
No. Speech is recognized on your computer by an open-source speech model that Koru downloads once. The microphone is opened by a separate small program, the only part of Koru that ever touches it, and it hands back words, not sound. Nothing is recorded or kept, and it works with no internet connection.
Can I dictate into other apps?
On Windows, yes. Turn on Koru Voice from the Voice input item in the notification area, and the same push-to-talk shortcut types into whatever app has focus: your email, a chat app, a browser form. Koru shows a small Listening pill near the bottom of the screen while it hears you, and the pill never steals focus.
For safety, Koru refuses to type into password and other protected fields, and if the window you were dictating into changes or disappears, it throws the words away rather than guess where they belong. On a Mac, Koru Voice works inside Koru today.
What won’t Koru Voice do?
It is deliberately careful:
- It does not type into code blocks, read-only documents or the code you enter when linking a device.
- Pressing Esc, clicking elsewhere before the words arrive, or the computer going to sleep ends the dictation without writing anything. Half a sentence in the wrong place is worse than losing it.
- Spoken punctuation is not interpreted yet: saying “period” types the word. Spoken commands for punctuation and new lines are planned.
How do I set it up?
- Download and open Koru.
- Press Ctrl+Shift+D (Cmd+Shift+D on a Mac) in a document. The first time, Koru offers to set up voice input and downloads its speech model.
- Choose your language, microphone and quality in Settings, or on Windows from the Voice input menu in the notification area.
- Hold the shortcut and talk.
While the model is loaded it uses several hundred megabytes of memory. The Voice input menu can release it from memory with one click; the next dictation loads it again after a short pause.
What’s next for voice in Koru?
Voice notes that keep the audio with your document, and meeting recording and transcription, are on the roadmap. Today’s Koru Voice turns speech into text and keeps nothing else.
Questions
Is Koru Voice free?
Yes. It is part of Koru, with no subscription, no account and no per-minute limits.
Does voice typing work offline?
Yes. The speech model runs on your computer, so dictation works with no internet connection once the model is downloaded.
Does Koru record or store my audio?
No. The microphone audio is turned into text on your computer and then discarded. Koru keeps only the words you choose to keep.
Can I use Koru Voice in Word, Outlook or Slack?
On Windows, yes: turn on Koru Voice from the notification area and the shortcut types into the app you are using. It will not type into password fields. On a Mac, Koru Voice works inside Koru.
Which languages does Koru Voice understand?
English, Spanish, Portuguese, German, French, Arabic, Chinese and Japanese. Choose the language in Settings.
Why is there a short pause before the first dictation?
Koru loads the speech model the first time you dictate, or after you released it from memory. Start dictation in the Voice input menu warms it up in advance.
Related
Last updated
